Mantis AI Harness™ · Industrial agentic AI
Mantis AI Harness spots problems early, explains the likely cause with evidence, and keeps the know-how of your experts in the company. Your data stays on site, and your people stay in control.
The platform
An industrial agentic AI platform that combines AI models with live plant data, your engineering documents and rule-driven reasoning, so its answers rest on what is happening in your operation, not on general knowledge.
Autonomous by design. Its agents watch your process around the clock, analyze what changes, plan their own reasoning steps and choose the tools they need, then alert the right person with the likely cause and the evidence. Every action that changes something still waits for a person's approval.
Each task is assigned to a model based on measured results: every call records the model used, its response time and its cost, so your team can see which model, local or cloud, serves each task best.
Responsible AI and trust. It runs on infrastructure you control, stays within your engineering and safety guardrails, asks a person before any write, cites the evidence behind each answer and flags unsupported or inconsistent claims.
The result: engineers and decision makers troubleshoot faster, understand complex processes, keep expert knowledge in the company, and speed up automation engineering.
Why it matters
Security and trust
Runs fully offline with local models.
No write without an operator's confirmation.
Cloud models only where you allow them.
Credentials never sit in manifests or logs.
Every action, approval and source recorded.
Each team sees only its own agents and data.
Your security team can review our security overview (PDF): Download it →
Industries we serve
Explain drops in throughput or recovery from plant data and your SOPs, with the evidence.
Watch boiler and plant tags, and flag abnormal combustion or steam conditions early.
Run power-flow studies on your network and explain the results in plain words.
Monitor pumps, levels and dosing, and explain what changed and why.
Trace batch deviations to their likely cause, with the data behind it.
Answer line questions from SOPs, drawings and cut sheets, citing the page.
Calculate Scope 2 footprints from utility bills with cited emission factors.
Track tank inventory and transfers at fuel terminals, and flag what is off plan.
Check structures against seismic design codes, and show the results in 2D and 3D.
How it works
People
Mantis kernel
On site or air-gapped · NVIDIA or Apple Silicon · or as a service
Specialist agents
Built from expert manifests
Autonomous reasoning
Plans its own steps, no scripted flows
Model routing
Right model per task
Governance
MCP open protocol
Plant data
OPC UA · historians
Knowledge
Documents · P&IDs
Simulation
Digital twin · power flow
Services
Any MCP server
Built with Claude
Mantis AI Harness was written with Claude Code, Anthropic's agentic coding tool, and runs Anthropic's Claude models alongside other providers wherever customers allow cloud models.
Try it
Inject a fault into a simulated boiler. The agent spots the drift, works out the likely cause, and proposes a fix your operator approves.
All values normal. The agent checks every tag continuously and alerts only when something drifts.
Runs in your browser. The plant, values and procedure are illustrative; a live demo on your own data is available on request.
From idea to agent
Goals, data sources, actions and limits, in plain language, with your experts.
The declarative blueprint: agent roles, knowledge, tools and guardrails.
The platform reads the manifest and assembles and wires the agents.
Agents run on your site, every action is audited, and the manifest evolves with the plant.
Start small, prove value
Phase 0 assesses your site and gets it ready: security review, network access, read-only data access, your documents and the agent. The 30 days of real use start only when Phase 0 is complete.
An appliance on your site with an annual license.
Appliance, updates and support for a monthly fee.
We first assess your site: network and OT security, OPC UA server and historian, documents, users and success criteria. Then we close the gaps: security approval, appliance installed, read-only data access, documents loaded, agent built with your experts.
Your users learn the interface and test agreed scenarios.
Real questions on real shifts, measured against an agreed baseline.
Results against the success criteria, and the next step.
Advisory
Deep practitioner experience with boardroom clarity, so enterprises capture the real value of AI and emerging technology.
Roadmaps for LLM deployment, agentic workflows, MLOps and AI governance, tied to business outcomes.
Sensors, edge gateways, SCADA integration, digital twins and analytics, designed for operational resilience.
Automation across the value chain, from scripted tasks to supervised AI workflows, with measured cycle-time and cost results.
Programs that change processes, technology and culture together, with clear milestones.
Multi-year plans, build-buy-partner decisions and board-ready investment cases.
Technology due diligence for investments and M&A, and provisional patent strategy for founders.
Research and proprietary IP
Technology developed in house, for partners who want to shape it early.
Reads P&IDs, ISA-88 recipes, I/O lists and interlock specifications, then generates IEC 61131-3 PLC code, SCADA configuration and test documentation, guided by a manifest that sets the standards and guardrails.
Continuous, evidence-based root-cause analysis across mining, oil and gas, distribution chains, bioprocessing and enterprise decision making: surfacing anomalies early, simulating corrections before committing, and sharpening fault catalogs over time.
One command interface for robot fleets: dispatch, patrol routing, health telemetry, safety interlocks and human override, each deployment defined by a manifest.
Organic emotional AI uses emotional modulation to shape attention, risk thresholds and arbitration in multi-agent systems before reasoning begins, so agents respond to people and situations the way experienced operators do.
About
Rayen Digital is an Ohio-based company with a team in Chile, built by engineers who have spent their careers putting AI and automation into real plants.
Decades in industrial automation, robotics, advanced manufacturing and industrial AI, including research in agentic AI for manufacturing and a portfolio of patents.
Mantis AI Harness, our platform for industrial AI agents, and new technology in root-cause analysis, code generation and human-AI interfaces.
Alongside your engineers and operators, starting with a focused pilot, and designing for safety, auditability and the realities of operational technology.
Insights
As AI moves into operations, every answer has to show its evidence: sources, calculations and a trail you can audit.
Read → Article · LinkedInWhy the hard part of industrial AI is getting it into daily operations, not making the models smarter.
Read → Post · LinkedInHow Rayen Digital keeps generative AI answers grounded before they reach production.
Read → Post · LinkedInHow declarative manifests are reshaping the way enterprises build and deploy AI agents.
Read → Post · LinkedInOne adaptive AI platform across conversational, physiological and industrial signals, with one governance artifact everyone can read.
Read →Contact
A confidential, no-obligation conversation about a pilot, an advisory project or a research partnership. NDA available on request.
Get the Mantis AI Harness brief (PDF)
Thank you. Your download should start now. If it does not, open the brief.